Release Strategies and the Social Impacts of Language Models
OpenAI；Harvard University（哈佛大学）；Cornell University（康奈尔大学）；Politiwatch；CTEC
英文摘要：
Large language models have a range of beneficial uses: they can assist in prose, poetry, and programming; analyze dataset biases; and more. However, their flexibility and generative capabilities also raise misuse concerns. This report discusses OpenAI's work related to the release of its GPT-2 language model. It discusses staged release, which allows time between model releases to conduct risk and benefit analyses as model sizes increased. It also discusses ongoing partnership-based research and provides recommendations for better coordination and responsible publication in AI.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
